Localization, engineered.
Kraken Intelligence runs your localization cycle end to end: configurable AI pipelines translate with your glossaries, translation memory and context; your linguists steer quality; costs, access and every change stay accounted for.
Access is provisioned per organization — sign in with SSO, Google, or your invitation.
Every string goes through the pipeline.
A translation is not a black-box call — it is a sequence of configurable, inspectable steps.
-
Source inapi · sdk · cms · files
Via the REST API, the Go SDK, the CMS module, or file upload; tasks fan out per target language.
-
Placeholder protectionplaceholder
Variables and markup are shielded before any engine sees the text.
-
Glossary enforcementglossary
Approved terms guide the AI; do-not-translate and forbidden terms are enforced deterministically.
-
Context retrievalcontext
Translation-memory matches and your knowledge context travel into the prompt with the string.
-
AI translate + post-editengine · post_edit
Your choice of engine — DeepL, Google, OpenAI, Anthropic, and more — with an AI post-edit pass.
-
Human review where neededhuman_review · lqa
Style and QA checks route tasks to review and LQA; linguists work in the TMS editor; nothing incomplete ships.
-
Deliveryapi · webhooks · write-back
Results return via the API and signed webhooks, and write back to the CMS and translation memory.
AI translates. Your team steers.
| SYS | System | Function | Fitted with |
|---|---|---|---|
| SYS-01 | Pipeline core | The pipeline is the engine. Every translation flows through composable steps you can see and change — not a black box. | visual editor · typed nodes · auto-selection · marketplace |
| SYS-02 | Context | Context beats raw MT. Glossaries, translation memories and knowledge collections feed every job, so output matches your terminology. | TM fuzzy/exact · TMX · TBX glossaries · image context |
| SYS-03 | Review | Humans steer quality. A full linguist TMS workflow — assignments, deadlines, a keyboard-driven editor, LQA — is built in, not bolted on. | assignments · TMS editor · human_review · LQA · net rates |
| SYS-04 | Feedback loop | The system learns from your edits. Post-edits are classified and become reviewable proposals — for prompts, glossary terms, QA rules. | post-edit classifier · prompt audit · proposal board · dashboards |
| SYS-05 | Tenancy | Trust is structural. Organization isolation, role-based access, audit, and transparent AI costs are product features, not plumbing. | RLS isolation · RBAC · audit log · budgets · cost per task |
| SYS-06 | Content | Content in any shape. Plain strings, key-value, JSON, files — and text localized directly on images. | CMS module · releases & rollback · staticloc images |
An API that ships with a manual.
- REST API. Submit content, poll tasks, upload files, manage CMS configs, run staticloc — camelCase JSON, async 202 responses.
- Signed webhooks. Standard Webhooks spec: HMAC-SHA256, zero-downtime secret rotation, byte-identical retries.
- Go SDK.
go get github.com/simple-life-app/localizetion-service/sdk - Live documentation. An interactive console at /docs/developers: assemble a real request and point every sample at your own host and key.
curl -sS -X POST \ "https://krakenloc.com/internal/api/v1/translate" \ -H "Authorization: Bearer dk_your_key" \ -H "Content-Type: application/json" \ -d '{ "sourceLang": "en", "targetLangs": ["ru", "de", "ja"], "items": [{ "type": "plain", "content": "Complete your purchase" }] }' // HTTP 202 → taskId · status: queued // result: GET /tasks/{id} or a signed webhook
Access for organizations.
Kraken Intelligence is provisioned per organization: your team signs in with SSO (Microsoft Entra ID), Google, or an email invitation. There is no public self-serve signup — if your organization has access, just log in.
Log in